Atlantic salmon (Salmo salar) COSEWIC assessment and status report: chapter 6

Wildlife Species Information

Name and Classification

Class: Osteichthyes / Actinopterygii
Order: Salmoniformes
Family: Salmonidae
Latin binomial: Salmo salar L.
Designatable Unit: See DU Section
Common species names:
English – Salmon, ouananiche (non-anadromous life history form)
French – Saumon atlantique
Other common names exist for various forms and life history stages of the species (e.g., see Froese and Pauly 2004).

Morphological Descriptioni

The most complete morphological description of Atlantic Salmon can be found in Scott and Crossman (1973) where it is described as having a ‘trout-like’ body with an average length of about 18 inches (457 mm), somewhat compressed laterally, with the greatest body depth usually at the dorsal fin origin or slightly posterior to it. The anadromous salmon has a blue-green back, silvery sides and a white belly (Carcao 1986). There are several X-shaped and round spots mostly above the lateral line (Carcao 1986). When a marine salmon re-enters freshwater it loses the silvery guanine coat replacing it with hues of greenish or reddish brown and large spots that are edged with white (Scott and Crossman 1973, Carcao 1986). Juvenile salmon, or parr, display ‘parr marks’ (pigmented vertical bands), with a single red spot between each parr mark along the lateral line (Scott and Crossman 1973). When parr are ready to migrate to sea, they are known as smolts. At this stage the parr marks are lost and the fish become silvery (Scott and Crossman 1973).

Spatial Population Structureii

A well-known characteristic of Atlantic Salmon is that mature adults generally return to their natal streams to spawn (recently reviewed in Hendry et al. 2004). However, some salmon do stray, spawn successfully, and produce offspring that are capable of surviving to spawn in later years. Analyses of molecular genetic variation can help determine the extent of reproductive isolation among salmon from different locations and hence the potential for adaptive differences to accrue (Waples 1991). Analyses of molecular genetic variation can also help identify highly divergent lineages that may have accumulated substantial genetic differences over long periods of reproductive isolation (Utter et al.1993).

A variety of studies of genetic variation within and among Atlantic Salmon populations have been carried out. Most have involved sample collections from several rivers from one or two regions, and a few have included collections from one or two rivers from several or all regions. These studies have all shown some degree of population structuring and genetic differentiation. They also suggest that individual rivers and in some cases even tributaries represent relatively independent demographic units.

The most informative genetic analysis of Atlantic Salmon populations in Quebec, New Brunswick and Labrador is that carried out by Dionne et al. (2008). Using a combination of landscape genetics and hierarchical analysis of genetic variance they identified seven regional groups (1: Ungava; 2: Labrador; 3: Lower North Shore; 4: Higher North Shore; 5: Quebec City; 6: Southern Quebec; 7: Anticosti; Figure 1) and showed that genetic variance among rivers within regions (2.02%) was less than variance among regions (2.54%). The extent of genetic differentiation among rivers from different regions was on average double that observed among rivers within any given region, although genetic differences between most pairs of rivers within regions were still statistically significant. Genetic divergence among populations and regions was correlated with coastal distance among rivers and degree of difference in temperature regime. In another study, Dionne et al. (2007) found that salmon appear to show some local adaptation in the form of genetic variation in MHC genes that is correlated with latitudinal changes in temperature regimes, which in turn are thought to drive clines in pathogen diversity.

Recent work in insular Newfoundland revealed genetic differentiation within rivers, primarily between anadromous and non-anadromous life history forms, but also among anadromous forms within relatively small watersheds (<1000 km2) (mean FST = 0.015-0.019, P < 0.05) for all pair-wise comparisons) (Adams 2007) (Figure 2). Adams (2007) did pair-wise comparisons of eight rivers in southern Labrador (Eagle River and south) and found a mean FST of 0.017 (P < 0.001). The divergence among rivers seemed to be influenced by river size. Divergence among several subsets of rivers (e.g., Alexis River and proximate rivers) was lower than expected, with no significant differences in multiple pair-wise comparisons. An examination of within-river structure by Dionne et al. (2009a) suggested significant within-river population structure. However, the degree was highly variable among rivers.

The influences of temporal variation, effective population size, life history variation, and local adaptation on gene flow among rivers and regions of Newfoundland and Labrador have also been examined (Palstra et al. 2007) (Figure 3). These authors demonstrated temporal stability across multiple generations and also suggested that metapopulation dynamics might be important in maintaining stability in smaller populations. Palstra et al. (2007) also suggested that the magnitude and directionality of gene flow among populations is variable and may even reverse direction when moving from contemporary to evolutionary time scales. Their work also suggested some level of correlation in life history and demographic attributes, and genetic population structure.

Verspoor (2005) reported that “variation among loci was highly heterogeneous at all polymorphic loci” for samples taken across Atlantic Canada, but did not provide information on specific pair-wise comparisons. King et al. (2001), in a hierarchical gene diversity analysis, partitioned variance among provinces or states, among rivers within provinces or states, and within rivers. The proportion of variance associated with among-river comparisons was 2.99% (within province or state), as opposed to 5.28% among countries in Europe. Pair-wise tests for significant differences among populations (rivers) were not provided. Bootstrap analyses were used by McConnell et al. (1997) to test for pair-wise differences among sample collections from different rivers for three different genetic distance measures, Roger’s modified genetic distance, allele sharing genetic distance, and Goldstein’s (δμ)2 distance. All pair-wise estimates of Roger’s distance and nearly all estimates of allele sharing genetic distance were significant, but very few estimates of Goldstein’s (δμ)2 distance were significant; most of these involved the Gander River, Newfoundland. Again, only a few rivers in each region were surveyed in this study.

Verspoor (2005) presented the most geographically comprehensive study published to date, and included multiple river populations from multiple regions (Newfoundland and Labrador, Quebec, Gulf, and Maritimes). In this study, variation was surveyed at 23 allozyme loci, of which 15 were informative (genetically variable). Multi-Dimensional Scaling analyses (Figure 4), and neighbour joining trees (Figure 5), both based on Nei’s DA distance, suggested the presence of six large-scale groupings of Atlantic Salmon in Eastern Canada: Labrador and Ungava, Gulf of Saint Lawrence, Newfoundland (excluding Gulf rivers), Atlantic Shore/Southern Upland of Nova Scotia, inner Bay of Fundy (iBoF), and outer Bay of Fundy (oBoF). Labrador and Ungava rivers grouped together, as did salmon from Newfoundland, excluding those from rivers that drain into the Gulf of Saint Lawrence. Generally speaking, salmon from the Atlantic coast of Nova Scotia (Southern Upland) clustered together and were distinct from all other samples analyzed, as were salmon from the inner Bay of Fundy. Many of the regional groupings identified above have also been reported in other studies, involving different molecular markers. Verspoor et al. (2002) identified an mtDNA haplotype in multiple inner Bay of Fundy rivers, at moderate to high frequency, that was completely absent in outer Bay of Fundy samples. In a recently expanded, though not yet published analysis of mtDNA in Atlantic salmon from Eastern Canada, Verspoor also noted the complete absence of the inner Bay mtDNA haplotype in 16 rivers of the Southern Upland. Verspoor et al. (2002) also identified an mtDNA haplotype in nearly all surveyed Southern Upland rivers that was absent in samples from all other surveyed salmon populations in Eastern Canada.

Spidle et al. (2003) and King et al. (2001), in surveys of variation in largely overlapping suites of microsatellites, found the inner Bay and Southern Upland populations included in the analysis to be highly distinct from all other populations analyzed (Figure 6). In a UPGMA tree of microsatellite-based pair-wise estimates of Roger’s genetic distance (McConnell et al. 1997), the 10 Southern Upland populations all clustered together, as did Stewiacke and St. Croix, NS populations (two inner Bay populations). The Gaspereau River again grouped separately from all other rivers, a likely result of a population bottleneck and rapid recent genetic drift.

Substantial evidence also exists for the distinctiveness of Newfoundland populations relative to other North American salmon populations in microsatellite allele (Spidle et al. 2003, King et al. 2001) and mtDNA haplotype (King et al. 2000) frequencies. Particularly notable are the presence of ‘European’ haplotypes in northeast coast Newfoundland populations, suggesting some post-glacial colonization of this area from European refugial populations.

Few surveys included samples from Labrador, and even fewer considered samples from Ungava (but see Fontaine et al. 1997 and Dionne et al. 2008). King et al. (2001) and Spidle et al. (2003) identified the Labrador populations as highly distinct from other populations. Adams (2007) compared samples from eight rivers in southern Labrador to four rivers from northeastern Newfoundland and found evidence of divergence at 10 microsatellite loci (FST = 0.021). The divergence, however, was similar to comparisons between insular Newfoundland rivers.

Non-genetic data support much of the broad-scale population structure inferred from the genetic data. For example, Chaput et al. (2006a) examined variation in life histories across the Canadian range of the species, including smolt age, small and large salmon proportions in returns, sea-age at maturity, proportion of small and large females, and fork length of small and large fish. This study was able to demonstrate clusters of populations with similar life history variation. For example, one clear differentiation was the dominance of grilse (one-sea-winter age at maturity) spawners in insular Newfoundland versus MSW-dominated populations in other areas. Populations also clustered based on smolt age and at-sea growth. Schaffer and Elson (1975) and Hutchings and Jones (1998) also demonstrated clear divergence in sea-age at maturity and size across regions.

Morphology and meristics have also been used to define salmon stocks in the North Atlantic. Claytor and MacCrimmon (1988) and Claytor et al. (1991) were able to show regional differentiation based on morphology, but meristic metrics were less successful. They concluded that insular Newfoundland, Labrador/Quebec, and the Maritime populations represented three very distinct regions. They also suggested, but with less certainty, that sub-structuring was likely in the Maritime regions.

Figure 1: Posterior Probabilities for each Atlantic Salmon River-specific Population Belonging to Each of the Seven Regional Groups in Quebec and Labrador Identified by Landscape Genetics Analysis

Charts showing posterior probabilities for each Atlantic Salmon river-specific population belonging to each of the seven regional groups in Quebec and Labrador.

The white area denotes a 90-100% probability that populations belong to their respective regional group. (a) Map of the river-specific populations included in the analysis. (b) Regional group 1: ‘Ungava’ (3 Rivers); (c) Regional group 2: ‘Labrador’ (7 rivers); (d) Regional group 3: ‘Lower North Shore’ 4 rivers); (e) Regional group 4: ‘Higher North Shore’ (10 rivers); (f) Regional group 5: ‘Quebec City’ (6 rivers); (g) Regional group 6: ‘Southern Quebec’ (18 rivers); (h) Regional group 7: ‘Anticosti’ (3 rivers) (Dionne et al. 2008).

Figure 2: Multidimensional Scaling Plot Based in Nei’s Unbiased Distance for Multiple Samples Taken from 4 Newfoundland Rivers and 8 Labrador Rivers

Multidimensional scaling plot based in Nei’s unbiased distance for multiple samples taken from four Newfoundland Rivers and eight Labrador rivers.

(1) Northwest River Salmon, (2) Northwest Pond ouananiche (non-anadromous form), (3) Endless Lake ouananiche, (4) Rocky River ouananiche Sample 1, (5) Rocky River salmon, (6) Rocky River smolt, (7) Little Salmonier River salmon, (8) Little Salmonier River juveniles, (9) Rocky River ouananiche sample 2, (10) Indian Bay Big Pond salmon, (11) Moccasin Pond ouananiche, (12) Wings Pond ouananiche, (13) Third Pond ouananiche, (14) Indian Bay Big Pond smolt, (15) Indian Bay Big Pond ouananiche, (16) Hungry Brook juveniles, (17) Eagle River, (18) Sandhill River, (19) St. Lewis River, (20) Alexis River, (21) Shinney’s Brook, (22) Black Bear River, (23) Paradise River, (24) Reed Brook (Adams 2007).

Figure 3: Multidimensional Acaling Plot for 20 Rivers in Newfoundland and Labrador, Using the First Two Dimensions that Capture 68% of the Genetic Variation

Multidimensional scaling plot for 20 rivers in Newfoundland and Labrador, using the first two dimensions that capture 68 percent of the genetic variation.

ENR English River, WAB Western Arm Brook, TNR Terra Nova River, MIB Middle Brook, GAR Gander River, FBB Flat Bay Brook, ROR Robinsons River, HLR Highland River, CRR Crabbes River, COR Conne River, SWB Southwest Brook, SMB Simmins Brook, BDN Baye Du Nord River, NWB Northwest Brook, NEB Northeast Brook, BBR Biscay Bay River, NEP Northeast River Placentia, NET Northeast Brook Trepassey, STR Stoney River (Palstra et al. 2007).

Figure 4: Allozyme Variation in Canadian Atlantic Salmon Populations

Three-part figure illustrating allozyme variation in Canadian Atlantic Salmon populations. Part A is a map showing the locations of 53 rivers that were included in a multilocus allozyme study. Part B is a list of the rivers. Part C is a multidimensional scaling plot for 48 rivers based on Nei’s DA genetic distance.

A, map showing locations of 53 rivers that were included in a multilocus allozyme study (Verspoor 2005). B, list of rivers. C, multidimensional scaling plot for 48 rivers based on Nei’s DA genetic distance. Large-scale groupings of Atlantic Salmon populations proposed by Verspoor (2005) are indicated. Modified from Verspoor (2005).

Figure 5: Neighbour-joining Dendrogram Based on Allozyme Data Using Nei’s Genetic Distance, for 48 Canadian Rivers

Neighbour-joining dendrogram based on allozyme data using Nei’s genetic distance, for 48 Canadian rivers.

(Verspoor 2005). See Figure 4 for regional groupings, river numbers are congruent.

Figure 6: Multidimensional Scaling Plot Based on Microsatellite Data for 16 Rivers in Canada

Multidimensional scaling plot based on microsatellite data for 16 rivers in Canada.

Canada (Newfoundland (NF), Quebec (QB), Nova Scotia (NS), New Brunswick (NB) and Maine (ME, MEL)). NF1 Conne, NF2 Gander, ME1,2,3,4 (Maine), NS1 Stewiacke, NS2 Gold, QB1 St. Jean, QB2 Saguenay, NB1 Naswaak, NB2 Miramichi, MEL1,2 (Maine Landlocked), LB1 Sandhill, LB2 Michaels (King et al. 2001).

Designatable Units

COSEWIC guidelines state that “a population or group of populations may be recognized as a DU if it has attributes that make it “discrete” and evolutionarily “significant” relative to other populations”. Evidence of discreteness can include “inherited traits (e.g. morphology, life history, behaviour) and/or neutral genetic markers (e.g. allozymes, DNA microsatellites…” as well as large disjunctions between populations, and occupation of different eco-geographic regions.

The well-known homing behaviour of Atlantic Salmon, as well as the morphological, life history, behavioural and molecular genetic data cited above, all indicate that the criterion of ‘discreteness’ is routinely satisfied at the level of rivers (as representative of discrete breeding populations), and indeed in some cases may be met at the level of tributaries within river drainages. Since Atlantic Salmon are believed to have spawned in ~700 rivers in Canada, this could suggest the possibility of a huge number of DUs; however, the second criterion of ‘evolutionary significance’ needs to be considered as well. The COSEWIC guidelines suggest four criteria for ‘significance’, three of which may be applicable to Atlantic Salmon.

The first ‘significance’ criterion is “evidence that the discrete population or group of populations differs markedly from others in genetic characteristics thought to reflect relatively deep intraspecific phylogenetic divergence”. This criterion is met for Atlantic Salmon at the ocean basin scale: a variety of molecular genetic data indicate that North American populations of Atlantic Salmon are divergent from European populations (e.g., King et al. 2000, 2001, Verspoor 2005). This deep split between eastern and western Atlantic Salmon populations is, however, of little relevance for assigning DUs of Canadian populations, except perhaps in one case. Atlantic Salmon populations in northeastern Newfoundland (DU 3, below) show the presence of ‘European’ mtDNA genotypes that do not naturally occur in any salmon populations to the south, suggesting that post-glacial colonization of this part of Newfoundland was in part from Europe (King et al. 2000). Apart from the mtDNA data for DU 3, there is little evidence of deep genetic distinctions (in neutral markers) among groups of Atlantic Salmon populations in Canada. The lack of evidence may in part be due to the relative lack of geographically comprehensive studies of genetic variation among Atlantic Salmon populations in Canada. Most studies have only sampled a portion of the Canadian range. The most geographically extensive genetic study to date is that of Verspoor (2005), which examined allozyme variation in 53 populations spanning most of the Canadian range. Verspoor (2005) suggested that the allozyme data supported the presence of six major population groups of salmon; however, the distinctions between groups were not large, and were not supported by statistical criteria (Figures 4 and 5).

The second ‘significance’ criterion of relevance is “persistence of the discrete population or group of populations in an ecological setting unusual or unique to the wildlife species, such that it is likely or known to have given rise to local adaptations”. As for discreteness, there is abundant evidence of varying local adaptations in Atlantic Salmon. Since Atlantic Salmon spend the first one to several years of their life in fresh water, many adaptations reflect local or regional variation in freshwater habitat attributes including, but not limited to, temperature, length of growing season, and pH. Other potentially adaptive variation includes variation among populations in the proportions of populations maturing as precocious male parr, or as one-sea-winter (1SW) or multi-sea-winter (MSW) salmon. Additional adaptive variation may include varying migration routes to distant ocean feeding grounds. At the molecular level, Dionne et al. (2007) found evidence of latitudinal clines in genetic variation at MHC loci, which they interpreted as evidence of adaptation to latitudinally varying assemblages of parasites.

Past attempts to artificially enhance local salmon populations by stocking them with hatchery-bred salmon derived from other populations have provided indirect evidence of local adaptation. For example, Ritter (1975) showed that the performance of hatchery-bred Atlantic Salmon stocked as smolts in rivers varied dramatically depending on the geographic distance between the ‘source’ populations (which were in the Gulf of St. Lawrence) and the ‘destination’ rivers in which they were stocked. Catches of salmon, both in distant marine fisheries and in local fisheries in or around the stocked river itself, were much lower when the salmon were stocked in rivers distant from the source rivers than when they were stocked in nearby rivers (Figure 7). Ritter (1975) concluded that the salmon did poorly when stocked outside their home region because of a mismatch between their adaptations and the locations in which they were stocked. Similarly, two reports on the status of Atlantic Salmon populations in Maine concluded that years of stocking of Maine rivers from several Canadian populations had not significantly eroded the genetic distinctiveness of a number of Atlantic Salmon populations in Maine, presumably because the stocked salmon were maladapted to local conditions (National Research Council 2002, 2004).

Figure 7: Recovery Rates for Stocked Atlantic Salmon Versus Distance from the Native River

Chart showing recovery rates for stocked Atlantic Salmon versus distance from the native river.

Shown are total recovery rates (both distant water ocean fisheries and in- or near-river terminal fisheries) for Atlantic Salmon stocked as smolts in rivers at varying distances from their native river. The results for distant water ocean fisheries and in- or near-river terminal fisheries are similar when analyzed separately (results not shown). Analysis of data from Ritter (1975) courtesy of C. Havie and P. O’Reilly.

The various lines of evidence cited above all indicate that Atlantic Salmon populations are locally adapted, and that they are therefore not ecologically exchangeable at some spatial scales. The difficulty lies in determining what those spatial scales are, or where differences among populations become great enough to merit status as DUs. Although it does not directly address this issue, the third COSEWIC ‘significance criterion of relevance to Atlantic Salmon may be of some help. It refers to “evidence that the loss of the discrete population or group of populations would result in an extensive gap in the range of the wildlife species in Canada”. Many of the DUs proposed below represent a sizable fraction of the species’ range in Canada, as well as showing some attributes of distinctiveness, and those DUs that are relatively small in area tend to have particularly strong evidence of genetic or ecological distinctiveness. It can be argued that the loss of any one of these units would represent a substantial loss of diversity within Atlantic Salmon in Canada.

Among the factors considered were genetic divergence, life history and morphometric variation, and geographic separation. As noted above, neutral genetic markers alone are not sufficient to define DUs, but they can, however, provide information on relative levels of gene flow among populations. Life history variation that was considered included data such as smolt age, sea age at maturity, run timing, migratory route, proportion female, and mean length at various life stages. Geographic separation was generally considered significant for major divisions such as insular Newfoundland versus mainland Canada, or north and south of the Gulf of St. Lawrence.

DU boundaries in Quebec and Labrador were guided in large part by the results of the extensive study conducted by Dionne et al. (2008). Using data from 13 microsatellite loci on salmon from 51 rivers, they used a combination of hierarchical and landscape genetic analyses in an effort to disentangle the relative influences of a range of factors (temperature, latitude, ‘coastal distance’ [from the southernmost population, the Miramichi], ‘migration tactic’ [shorter migrating 1SW vs. longer migrating MSW salmon], an index of the ‘difficulty of upstream migration’, and stocking history) on genetic structure of Atlantic Salmon populations in the Quebec-Labrador region. They identified seven regional groupings of Atlantic Salmon, which have been adopted as DUs. Temperature and distance, both between rivers and from the southern boundary of the study area, emerged as key determinants of the genetic structure of Atlantic Salmon populations. The influence of distance from the south was suggested to be the “historical footprint of the North American colonization process” from a glacial refugium southward of the contemporary range. In other words, historical effects dating from early post-glacial colonization remain evident in contemporary population structure. Importantly, evidence of dispersal was detected, both within and among population groupings, but genetic differentiation between rivers was lower for dispersal within population groups than it was for similar levels of dispersal between population groups. This observation led the authors to hypothesize that gene flow (as opposed to dispersal) between population groups is constrained by differing thermal regimes which promote local adaptation within groups.

The Department of Fisheries and Oceans (DFO) has previously defined 28 Conservation Units (CUs) for Atlantic Salmon (DFO and MRNF 2008; Figure 8); whereas, 16 DUs are recognized (Figure 9). Despite the difference in the numbers of DUs and CUs, and the fact that the DUs were developed independently, the 16 DUs share many features with the 28 CUs. The majority of boundaries between DUs coincide with CU boundaries. Nine DUs (1, 3, 5, 6, 9, 11, 14, 15, 16) correspond to (differently numbered) CUs. Two DUs (4, 13) each comprise two CUs. One DU (2) combines two very large and one very small CU in Labrador, and unlike the CUs, extends into Quebec. Three DUs within Quebec have different boundaries than the CUs in the same area and together include five CUs and parts of two others. DU 12 (Gaspé-Southern Gulf of St. Lawrence) comprises all of six CUs, and part of another. The similarities between DUs and CUs reflects the similarity of the definition used for CUs (“groups of individuals likely exhibiting unique adaptations that are largely reproductively isolated from other groups, and that may represent an important component of a species’ biodiversity”; DFO and MRNF 2008) to the criteria used by COSEWIC to recognize DUs. The differences largely reflect two factors: the availability of newer data, particularly those in Dionne et al. (2008), which formed the basis for decisions about DU structure in the Quebec-Labrador region, and an operational strategy of lumping CUs within DUs when evidence supporting splitting was judged to be weak. The relatively large DU 2 (Labrador) and DU 12 (Gaspé – Southern Gulf of St. Lawrence) reflect this strategy of lumping CUs in the absence of strong data for splitting. The structure for these large DUs may require refinement in the future as more data become available. In the following descriptions, DUs are cross-referenced with DFO CUs and Salmon Fishing Areas, and Quebec Fishing Zones. A tabular comparison of DU characteristics is presented in Table 1.

Figure 8: Conservation Units (CUs) Proposed by the Department of Fisheries and Oceans for Atlantic Salmon

Map showing locations of conservation units proposed by the Department of Fisheries and Oceans for Atlantic Salmon.

(DFO and MRNF 2008).

Figure 9: Proposed Designatable Units (DU) for Atlantic Salmon in Eastern Canada

Map showing the locations of proposed designatable units for Atlantic Salmon in eastern Canada.


i This section is taken from COSEWIC 2006a.
ii Elements of this section are copied, abstracted and/or synthesized from DFO and MRNF (2008).

Report a problem or mistake on this page
Please select all that apply:

Thank you for your help!

You will not receive a reply. For enquiries, contact us.

Date modified: